Increased sensitivity in identifying language-related functional connectivity using jackknife resampling analyses

Abstract Functional connectivity (FC) analyses of task-based fMRI (tbfMRI) often rely on static correlation methods that average signal relationships over time. While widely used, these methods may miss transient but meaningful neural interactions. In this study, we investigated whether jackknife resampling—a technique that systematically omits one time point at a time—enhances sensitivity in detecting language-related FC networks during an auditory comprehension task. We analyzed surface-based FC networks in 172 healthy young adults from the Human Connectome Project. FC matrices were computed across 68 cortical regions of interest, and statistically significant edges were identified using Bonferroni correction. We compared FC networks derived from a traditional static correlation approach with those obtained using jackknife resampling, applying an edge consistency threshold to retain only the most stable connections across time points. The static method identified 75 significant language-related FCs. Jackknife-based analyses recovered all of these and revealed 24 additional connections or edges (eight left-hemispheric, five right-hemispheric, 11 interhemispheric; p < 0.001), including well-established language regions such as the middle temporal gyrus and posterior cingulate cortex. Jackknife resampling enhances detection of robust, task-relevant FCs, offering a promising alternative for modeling language networks and improving neurocomputational representations in both research and clinical settings.